In this work we couple SSL and MIL to train a deep learning classifier that combines the advantages of both methods and overcomes their limitations.
For me, computational pathology was a completely new field which yields some very specific challenges for machine learning algorithms – Whole Slide Images have very high resolutions and labeled data is often scarce. In this article we provide a possible solution for these challenges by coupling multiple instance and semi supervised learning. This allows us to train a machine learning model with a small percentage of labeled data while still maintaining a high performance. To know more, please check out the article https://ieeexplore.ieee.org/document/9681818.
The perspective to actually help people with better diagnostic tools is a huge motivation and I am curious about everything yet to come.
Proposed training framework, combining semi supervised and multiple instance learning for efficient cancer classification. It utilizes the global WSI diagnosis, unlabeled patches and a limited amount of labeled patches for training. The model requires less labels while showing a performance comparable to supervised state-of-the-art methods.
Thanks for the great collaboration to Julio Silva Rodríguez, Valery Naranjo Ornedo and my supervisor Rafael Molina
Arne Schmidt – ESR8