EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

LTM-UNet: Linear Transformer–Mamba with Attention-Based U-Net for Context-Aware Breast Ultrasound Image Segmentation

Shivpratap Singh Kushwah iD, Santosh Prakash Chouhan iD, Narinder Singh Punn iD, Mahua Bhattacharya iD

Diagnostics · 17.06.2026 · 16 (12) · 1888

DOI: 10.3390/diagnostics16121888 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Diagnostics, online erschienen am 17.06.2026. Verfasst von Shivpratap Singh Kushwah und 3 weiteren Autor:innen. Beteiligte Einrichtung: Department of Information Technology, Atal Bihari Vajpayee Indian Institute of Information Technology and Management, Gwalior 474015, MP, India. Die Arbeit nennt 49 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background/Objectives: Accurate breast lesion segmentation using deep learning models requires precise understanding of both global contextual relevance and finer lesion structure details, which remains a challenge for existing convolutional and transformer-based approaches. This study aims to address these limitations by proposing a new segmentation model capable of improving context-aware dense segmentation tasks for ultrasound images.

Method

We propose LTM-UNet, a novel segmentation method integrating transformer-based encoding with state-space-driven decoding in a U-Net-style framework. The architecture utilizes an efficient vision transformer encoder to extract multi-scale global representations. These features are refined through an attention-guided skip-fusion mechanism incorporating spatial-channel attention preserving finer spatial details and thereby minimizes the semantic gap between encoder and decoder features. Additionally, a direction-aware decoder based on a state-space model is introduced to efficiently capture long-range dependencies and enhance relevant feature reconstruction.

Results

Extensive experiments on benchmark ultrasound medical imaging datasets demonstrate the effectiveness of the proposed method. The model achieves dice-score coefficients of 82.41% on the BUSI dataset and 86.62% on Dataset B (UDIAT), outperforming several existing segmentation approaches in both dice-score coefficient and Intersection-over-Union (IoU) metrics.

Conclusions

The integration of efficient transformer-based global feature extraction, attention-enhanced feature fusion, and state-space-driven decoding enables LTM-UNet to effectively capture both structural details and contextual information, resulting in superior segmentation performance compared to existing methods.

Schlagwörter

Stichwörter der Autor:innen

attention-guided skip fusionbreast lesionstate-space modeltransformerultrasound medical imaging

Autor:innen und Institutionen

  • Department of Information Technology, Atal Bihari Vajpayee Indian Institute of Information Technology and Management, Gwalior 474015, MP, India.
  • Centre for Biomedical Research, Atal Bihari Vajpayee Indian Institute of Information Technology and Management, Gwalior 474015, MP, India.

ORCID-Profile: Shivpratap Singh Kushwah, Santosh Prakash Chouhan, Narinder Singh Punn, Mahua Bhattacharya

Zitieren

APA

Kushwah, S. S., Chouhan, S. P., Punn, N. S., & Bhattacharya, M. (2026). LTM-UNet: Linear Transformer–Mamba with Attention-Based U-Net for Context-Aware Breast Ultrasound Image Segmentation. Diagnostics, 16(12), 1888. https://doi.org/10.3390/diagnostics16121888

Vancouver

Kushwah SS, Chouhan SP, Punn NS, Bhattacharya M. LTM-UNet: Linear Transformer–Mamba with Attention-Based U-Net for Context-Aware Breast Ultrasound Image Segmentation. Diagnostics. 2026;16(12):1888. doi:10.3390/diagnostics16121888

BibTeX
@article{Kushwah2026,
  author = {Kushwah, Shivpratap Singh and Chouhan, Santosh Prakash and Punn, Narinder Singh and Bhattacharya, Mahua},
  title = {LTM-UNet: Linear Transformer–Mamba with Attention-Based U-Net for Context-Aware Breast Ultrasound Image Segmentation},
  journal = {Diagnostics},
  year = {2026},
  volume = {16},
  number = {12},
  pages = {1888},
  doi = {10.3390/diagnostics16121888}
}
Kostenlos starten →