Skip to content
DraconiticResearch
Search Sources API
197,768 studies31 sourcesUpdated 53 min ago
Peer-reviewedCC BY 4.0PLOS One

LENS: A mammography-specific hybrid CNN-Transformer with lesion-aware evidence modeling

  • Duc Quy Hoang1
  • Van Kien Cao2
  • Tan Nhu Nguyen3
  • Ngoc Son Nguyen4
4 affiliations
  1. 1Faculty of Electronics Technology, Industrial University of Ho Chi Minh City, Ho Chi Minh City, Viet Nam
  2. 2Big Data Research for Infrastructure and Green Engineering (BRIDGE) Research Group, Ho Chi Minh City University of Transport, Ho Chi Minh City, Viet Nam
  3. 3School of Biomedical Engineering, International University, Ho Chi Minh City, Viet Nam
  4. 4Vietnam National University, Ho Chi Minh City, Viet Nam
Open at sourcePDFDOICite

Abstract

Background: Artificial Intelligence (AI) models for mammography classification is prone to shortcut learning because diagnostically relevant evidence is typically sparse, localized, and easily dominated by non-lesion background context. This study aimed to develop a mammography-specific framework that integrates lesion-focused evidence with global image representations to improve classification performance and provide more clinically interpretable decision support. Methods: We propose LENS, a hybrid CNN-Transformer family designed specifically for mammography. LENS combines a lightweight multi-scale convolutional neural network (CNN) backbone with an alternating local-global Transformer encoder. In addition to global image representations, a weakly supervised lesion-aware branch identifies and aggregates suspicious regional evidence to support image-level prediction. LENS was evaluated on the large-scale VinDrMammo dataset for the three-class mammography task and compared to advanced architectures, including ConvNeXt, DINOv2, GMIC, and Swin Transformer under a unified experimental protocol. Results: LENS-Base achieved the highest overall Accuracy of 85.5%, Macro F1-score of 79.6%, and Matthews correlation coefficient of 0.65. Quantitative localization evaluation results indicated that the selected regional evidence frequently overlapped with annotated abnormalities. Furthermore, these promising results come with fewer parameters and lower FLOPs. Conclusion: LENS integrates local lesion-related features with global anatomical context to achieve improved class-balanced mammography classification while providing quantitatively supported lesion-focused evidence. These findings suggest that LENS is a promising framework for AI-assisted mammography screening.

Related

  1. Hybrid local–global vision transformer framework for robust plant disease detection in precision agriculture
    Anshu Sharma, Chirag Sharma, Ramandeep Sandhu +2
    Scientific ReportsPeer-reviewedCC BY-NC-ND 4.0Full textPlant biologyCrossref
  2. A hybrid EfficientNet-B0-CBAM-Vision Transformer framework for accurate cherry leaf disease classification
    Ateeq Ur Rehman, Jaeyoung Choi, Ahmad Almogren +3
    Scientific ReportsPeer-reviewedCC BY-NC-ND 4.0Plant biologyCrossref
  3. UniHAR: A unified hierarchical transformer with multi-scale temporal modeling for wearable human activity recognition
    Memoona, Muhammad Haseeb Hashir, Sung Won Kim
    PLOS OnePeer-reviewedCC BYFull textNeuroscienceStatisticsPLOSEurope PMCPMC
  4. A linear-transformer hybrid for single-nucleotide polymorphism-based genotype-to-phenotype prediction in grapevine.
    Yibin Wang, Murukarthick Jayakodi, Silvas Kirubakaran +2
    The plant genomePeer-reviewedCC BYFull textGenetics & genomicsEurope PMCPMC
  5. Transformer-aided guard: A hybrid spatio-temporal framework for confidence-aware intrusion detection in cyber-physical systems
    Abeer Aljohani
    PLOS OnePeer-reviewedCC BY 4.0Full textNeuroscienceGeneral mathematicsPLOSPMC

Cite

BibTeXRIS
BibTeX
@article{hoang2026lens,
  title = {{LENS: A mammography-specific hybrid CNN-Transformer with lesion-aware evidence modeling}},
  author = {Duc Quy Hoang and Van Kien Cao and Tan Nhu Nguyen and Ngoc Son Nguyen},
  journal = {PLOS One},
  year = {2026},
  month = sep,
  doi = {10.1371/journal.pone.0350720},
  url = {https://doi.org/10.1371/journal.pone.0350720},
  note = {CC BY 4.0},
}
RIS
TY  - JOUR
TI  - LENS: A mammography-specific hybrid CNN-Transformer with lesion-aware evidence modeling
AU  - Duc Quy Hoang
AU  - Van Kien Cao
AU  - Tan Nhu Nguyen
AU  - Ngoc Son Nguyen
PY  - 2026
DA  - 2026/09/01
JO  - PLOS One
DO  - 10.1371/journal.pone.0350720
UR  - https://doi.org/10.1371/journal.pone.0350720
AB  - Background: Artificial Intelligence (AI) models for mammography classification is prone to shortcut learning because diagnostically relevant evidence is typically sparse, localized, and easily dominated by non-lesion background context. This study aimed to develop a mammography-specific framework that integrates lesion-focused evidence with global image representations to improve classification performance and provide more clinically interpretable decision support. Methods: We propose LENS, a hybrid CNN-Transformer family designed specifically for mammography. LENS combines a lightweight multi-scale convolutional neural network (CNN) backbone with an alternating local-global Transformer encoder. In addition to global image representations, a weakly supervised lesion-aware branch identifies and aggregates suspicious regional evidence to support image-level prediction. LENS was evaluated on the large-scale VinDrMammo dataset for the three-class mammography task and compared to advanced architectures, including ConvNeXt, DINOv2, GMIC, and Swin Transformer under a unified experimental protocol. Results: LENS-Base achieved the highest overall Accuracy of 85.5%, Macro F1-score of 79.6%, and Matthews correlation coefficient of 0.65. Quantitative localization evaluation results indicated that the selected regional evidence frequently overlapped with annotated abnormalities. Furthermore, these promising results come with fewer parameters and lower FLOPs. Conclusion: LENS integrates local lesion-related features with global anatomical context to achieve improved class-balanced mammography classification while providing quantitatively supported lesion-focused evidence. These findings suggest that LENS is a promising framework for AI-assisted mammography screening.
SN  - 1932-6203
ER  -