EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Noisy Teacher-Student Knowledge Distillation for Efficient Mass Detection in Mammography

Khai-Thinh Nguyen, Bach-Tung Pham, Chanh-Nghiem Nguyen, Thi-Phuong Le, Huang-Chia Shih, Yi-Chiung Hsu, Yung-Hui Li, Chien-Yao Wang, Jia-Ching Wang

IEEE Access · 2026 · 14 · 90386-90398

DOI: 10.1109/access.2026.3700401 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Khai-Thinh Nguyen und 8 weiteren Autor:innen. Die Arbeit nennt 62 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Breast cancer screening plays a vital role in reducing the mortality rate due to breast cancer. However, this process is not always accessible in rural areas. In recent years, deep-learning-based computer-aided diagnosis (CAD) systems for breast cancer detection have been developing, but developing these systems still remains a challenge due to limited annotated datasets. To alleviate this shortcoming, we proposed Lateral Auxiliary Augmentation (LAA), which combines a normal image with a mass-containing image, enabling a model to learn features from two different domains simultaneously. Then, the proposed noisy teacher–student knowledge distillation was applied to train a lightweight student model with knowledge from the larger teacher model. The best student model had AP:50 scores of 0.641 and 0.826 on the CBIS-DDSM test set and the independent INBreast dataset, respectively. Additionally, on the INBreast dataset, this model also achieved a true positive rate (TPR) of 0.93 at a false positive per image (FPI) of 0.32. These results demonstrated that the proposed distillation framework allowed training a lightweight model that achieved competitive results in mass detection. The lightweight model can be deployed on affordable computing devices, enabling CAD for mammography screening in areas where powerful computing devices are unavailable.

Zitieren

APA

Nguyen, K., Pham, B., Nguyen, C., Le, T., Shih, H., Hsu, Y., Li, Y., Wang, C., & Wang, J. (2026). Noisy Teacher-Student Knowledge Distillation for Efficient Mass Detection in Mammography. IEEE Access, 14, 90386-90398. https://doi.org/10.1109/access.2026.3700401

Vancouver

Nguyen K, Pham B, Nguyen C, Le T, Shih H, Hsu Y, et al. Noisy Teacher-Student Knowledge Distillation for Efficient Mass Detection in Mammography. IEEE Access. 2026;14:90386-90398. doi:10.1109/access.2026.3700401

BibTeX
@article{Nguyen2026,
  author = {Nguyen, Khai-Thinh and Pham, Bach-Tung and Nguyen, Chanh-Nghiem and Le, Thi-Phuong and Shih, Huang-Chia and Hsu, Yi-Chiung and Li, Yung-Hui and Wang, Chien-Yao and Wang, Jia-Ching},
  title = {Noisy Teacher-Student Knowledge Distillation for Efficient Mass Detection in Mammography},
  journal = {IEEE Access},
  year = {2026},
  volume = {14},
  pages = {90386--90398},
  doi = {10.1109/access.2026.3700401}
}
Kostenlos starten →