Explainable Feature Learning with Variational Autoencoders for Holographic Image Analysis

Stefan Röhrl, Lukas Bernhard, Manuel Lengl, Christian Klenk, Dominik Heim, Martin Knopp, Martin Knopp, Simon Schumann, Oliver Hayden, Klaus Diepold

2023

Abstract

Digital holographic microscopy (DHM) has a high potential to be a new platform technology for medical diagnostics on a cellular level. The resulting quantitative phase images of label-free cells, however, are widely unfamiliar to the bio-medical community and lack in their degree of detail compared to conventionally stained microscope images. Currently, this problem is addressed using machine learning with opaque end-to-end models or inadequate handcrafted morphological features of the cells. In this work we present a modified version of the variational Autoencoder (VAE) to provide a more transparent and interpretable access to the quantitative phase representation of cells, their distribution and their classification. We can show a satisfying performance in the presented hematological use cases compared to classical VAEs or morphological features.

Paper Citation

in Harvard Style

Röhrl S., Bernhard L., Lengl M., Klenk C., Heim D., Knopp M., Schumann S., Hayden O. and Diepold K. (2023). Explainable Feature Learning with Variational Autoencoders for Holographic Image Analysis. In Proceedings of the 16th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2023) - Volume 2: BIOIMAGING; ISBN 978-989-758-631-6, SciTePress, pages 69-77. DOI: 10.5220/0011632800003414

in Bibtex Style

@conference{bioimaging23,
author={Stefan Röhrl and Lukas Bernhard and Manuel Lengl and Christian Klenk and Dominik Heim and Martin Knopp and Simon Schumann and Oliver Hayden and Klaus Diepold},
title={Explainable Feature Learning with Variational Autoencoders for Holographic Image Analysis},
booktitle={Proceedings of the 16th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2023) - Volume 2: BIOIMAGING},
year={2023},
pages={69-77},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0011632800003414},
isbn={978-989-758-631-6},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 16th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2023) - Volume 2: BIOIMAGING
TI - Explainable Feature Learning with Variational Autoencoders for Holographic Image Analysis
SN - 978-989-758-631-6
AU - Röhrl S.
AU - Bernhard L.
AU - Lengl M.
AU - Klenk C.
AU - Heim D.
AU - Knopp M.
AU - Schumann S.
AU - Hayden O.
AU - Diepold K.
PY - 2023
SP - 69
EP - 77
DO - 10.5220/0011632800003414
PB - SciTePress