Disentangled representations of microscopy images
Keywords: Microscopy images, Disentangled representations, Transfer learning, Interpretability
TL;DR: We transfer a disentangled representation from a Source dataset to real-world Target dataset reaching a compromise between classification accuracy of the downstream task and interpretability in microscopy images
Abstract: Microscopy image analysis is fundamental for different applications, from diagnosis to synthetic engineering and environmental monitoring. In the last few years, the number of available images has been constantly growing, thanks to technological advancements, pushing toward the development of automatic image analysis methods based on deep learning. Although deep neural networks have demonstrated great performance in this field, interpretability — an essential requirement for microscopy image analysis — remains an open challenge.
This work proposes a Disentangled Representation Learning (DRL) methodology to enhance model interpretability for microscopy image classification.
Exploiting benchmark datasets coming from three different microscopic image domains, including plankton, yeast vacuoles, and human cells, we show how a DRL framework, based on transfer learning from synthetic features, can provide a good trade-off between accuracy and interpretability in this domain.
Supplementary Material: zip
Primary Area: applications to physical sciences (physics, chemistry, biology, etc.)
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 11420
Loading