A Systematic Review: The Application of Attention Mechanisms in Medical Ultrasound Image Processing

Clinical Snapshot

Key Findings

• P Value: Not reported at the review level

• Effect Size: Not reported — no pooled effect size calculated; individual study results are summarised qualitatively

• Primary Outcome: Narrative summary of attention mechanism applications in ultrasound image segmentation, detection, and classification across multiple organ systems, categorised by attention type (channel, spatial, hybrid)

• Nnt Or Sensitivity: Not reported — no diagnostic accuracy meta-analysis performed; individual study sensitivity/specificity values are not pooled or summarised with precision estimates

• Confidence Interval: Not reported — no quantitative synthesis performed

Clinical Application

Implementation of attention mechanism-based ultrasound AI tools in clinical practice requires substantial additional steps beyond what this review addresses: prospective multi-centre validation, regulatory approval, integration with existing PACS and reporting workflows, clinician training, and ongoing performance monitoring. The review does not address any of these translational requirements. Computational infrastructure requirements for real-time inference may also be a barrier in resource-limited settings. In Australia, AI-based medical imaging tools require TGA approval as Software as a Medical Device (SaMD) under the Therapeutic Goods Act 1989, classified according to the IMDRF risk framework. No specific attention mechanism-based ultrasound AI tools are referenced in this review as having received TGA listing. The Royal Australian and New Zealand College of Radiologists (RANZCR) and ASUM (Australasian Society for Ultrasound in Medicine) have published position statements on AI in medical imaging emphasising the need for prospective validation, transparency, and clinician oversight — standards this review's underlying evidence base does not yet meet. Medicare Benefits Schedule (MBS) reimbursement for AI-assisted ultrasound interpretation remains undefined. Australian clinicians should treat this review as horizon-scanning literature rather than practice-guiding evidence. Potentially applicable to clinical settings where ultrasound is used for diagnostic imaging across thyroid, breast, cardiac, obstetric, abdominal, and musculoskeletal applications. However, applicability is currently limited to research and development contexts rather than direct clinical deployment, given the absence of prospective validation data and regulatory approval for specific attention-based systems.

Abstract

Among numerous medical imaging modalities, ultrasound imaging is one of the most commonly used diagnostic methods in clinical practice. However, ultrasound diagnosis heavily relies on physician experience, and diagnostic results often lack reproducibility. In recent years, with the rapid development of artificial intelligence technology, which provides new impetus for the automated medical ultrasound image processing and analysis. Among the numerous deep learning approaches proposed, attention mechanisms have become a key component for improving network robustness to cope with challenges in ultrasound imaging, such as low contrast, blurred boundaries, and variable object morphologies. This paper systematically reviews the attention mechanisms employed in medical ultrasound image analysis, which can be roughly divided these mechanisms into three categories based on the differences in feature focus dimensions: channel attention, spatial attention, and hybrid attention. Most importantly, we not only summarized the application scenarios and effectiveness of various attention mechanisms but also analyzed the potential challenges faced in the future.

References

1. 1.Feng, W., Sun, S., Xue, Z., Shi, Y., & Chen, G. (2026). A systematic review: The application of attention mechanisms in medical ultrasound image processing. Ultrasound in Medicine & Biology. Advance online publication. https://doi.org/10.1016/j.ultrasmedbio.2026.03.002