EUVIMEDEUVIMED entdecken
Crossref

FoPro-KD: Fourier Prompted Effective Knowledge Distillation for Long-Tailed Medical Image Recognition

Marawan Elbatel, Robert Marti, Xiaomeng Li

IEEE Transactions on Medical Imaging · 2024-03 · 43 (3) · 954-965

DOI: 10.1109/tmi.2023.3327428 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Medical Imaging, 2024-03. Verfasst von Marawan Elbatel und 2 weiteren Autor:innen. Hauptthemen laut MEDLINE: Rare Diseases, Data Compression. Die Arbeit wurde bisher 20-mal zitiert (Crossref) und nennt 40 Literaturangaben. Gefördert von: Foshan Hong Kong University of Science and Technology (HKUST) Projects, Project of Hetao Shenzhen-Hong Kong Science and Technology Innovation Cooperation Zone, European Education and Culture Executive Agency (EACEA) Erasmus Mundus Grant.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Representational transfer from publicly available models is a promising technique for improving medical image classification, especially in long-tailed datasets with rare diseases. However, existing methods often overlook the frequency-dependent behavior of these models, thereby limiting their effectiveness in transferring representations and generalizations to rare diseases. In this paper, we propose FoPro-KD, a novel framework that leverages the power of frequency patterns learned from frozen pre-trained models to enhance their transferability and compression, presenting a few unique insights: 1) We demonstrate that leveraging representations from publicly available pre-trained models can substantially improve performance, specifically for rare classes, even when utilizing representations from a smaller pre-trained model. 2) We observe that pre-trained models exhibit frequency preferences, which we explore using our proposed Fourier Prompt Generator (FPG), allowing us to manipulate specific frequencies in the input image, enhancing the discriminative representational transfer. 3) By amplifying or diminishing these frequencies in the input image, we enable Effective Knowledge Distillation (EKD). EKD facilitates the transfer of knowledge from pre-trained models to smaller models. Through extensive experiments in long-tailed gastrointestinal image recognition and skin lesion classification, where rare diseases are prevalent, our FoPro-KD framework outperforms existing methods, enabling more accessible medical models for rare disease classification. Code is available at https://github.com/xmed-lab/FoPro-KD.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Rare DiseasesData CompressionHumansSkin

Förderung

  • Foshan Hong Kong University of Science and Technology (HKUST) Projects
  • Project of Hetao Shenzhen-Hong Kong Science and Technology Innovation Cooperation Zone
  • European Education and Culture Executive Agency (EACEA) Erasmus Mundus Grant
  • Spanish Science and Innovation Ministry, conducted with joint supervision of Robert Martí and Xiaomeng Li

Zitieren

APA

Elbatel, M., Marti, R., & Li, X. (2024). FoPro-KD: Fourier Prompted Effective Knowledge Distillation for Long-Tailed Medical Image Recognition. IEEE Transactions on Medical Imaging, 43(3), 954-965. https://doi.org/10.1109/tmi.2023.3327428

Vancouver

Elbatel M, Marti R, Li X. FoPro-KD: Fourier Prompted Effective Knowledge Distillation for Long-Tailed Medical Image Recognition. IEEE Transactions on Medical Imaging. 2024;43(3):954-965. doi:10.1109/tmi.2023.3327428

BibTeX
@article{Elbatel2024,
  author = {Elbatel, Marawan and Marti, Robert and Li, Xiaomeng},
  title = {FoPro-KD: Fourier Prompted Effective Knowledge Distillation for Long-Tailed Medical Image Recognition},
  journal = {IEEE Transactions on Medical Imaging},
  year = {2024},
  volume = {43},
  number = {3},
  pages = {954--965},
  doi = {10.1109/tmi.2023.3327428}
}
Kostenlos starten →