Skip to content

This project focuses on the VisioMel Challenge whose goal is predicting melanoma relapse. Recent advancements in SSL and WSL offer promising new solutions for improving the accuracy of cancer relapse detection.

License

Notifications You must be signed in to change notification settings

souheib1/Mixing-weakly-supervised-and-self-supervised-learning-techniques-for-melanoma-relapse-detection

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

24 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Mixing weakly-supervised and self supervised-learning techniques for melanoma relapse detection

This project focuses on the VisioMel Challenge whose goal is predicting melanoma relapse. It is a complex and often unreliable task, with diagnostic accuracy often varying among experienced medical professionals. Recent advancements in SSL and WSL offer promising new solutions for improving the accuracy of cancer relapse detection.

Capture

In this work, we specifically explore combinations of SSL and WSL techniques, namely BYOL (Bootstrap Your Own Latent) followed by AbMIL (Attention-based Multiple Instance Learning) and SimCLR (Simple Framework for Contrastive Learning of Representations) followed by AbMIL. BYOL and SimCLR are SSL techniques that train models using unlabeled data, while AbMIL is a WSL approach that exploits weakly labeled data without pixel-level annotations.

By leveraging the strengths of BYOL and SimCLR for self-supervised learning, as well as AbMIL for weakly supervised learning, our research provides insight into the effectiveness of using these techniques in the context of melanoma relapse prediction.

PIPELINE

References

[1] Ting Chen et al. “A Simple Framework for Contrastive Learning of Visual Representations”. In: ArXiv abs/2002.05709 (2020).

[2] Jean-Bastien Grill et al. “Bootstrap your own latent-a new approach to self-supervised learning”. In: Advances in neural information processing systems 33 (2020), pp. 21271–21284. Link to the paper

[3] Maximilian Ilse, Jakub M. Tomczak, and Max Welling. Attention-based deep multiple instance learning. June 2018. Link to the paper

[4] Bin Li, Yin Li, and Kevin W. Eliceiri. Dual-stream multiple instance learning network for whole slide image classification with self-supervised contrastive learning. Apr. 2021. Link to the paper

[5] Ming Y. Lu et al. Data efficient and weakly supervised computational pathology on whole slide images. May 2020. Link to the paper

About

This project focuses on the VisioMel Challenge whose goal is predicting melanoma relapse. Recent advancements in SSL and WSL offer promising new solutions for improving the accuracy of cancer relapse detection.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published