Automated detection of macrophages in quantitative phase images by deep learning using a Mask Region-based Convolutional Neural Network

Kai Eder, Tobias Kutscher, Anne Marzi, Álvaro Barroso, Jürgen Schnekenburger, Björn Kemper

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Citation (Scopus)

Abstract

We explored a Mask Region-based Convolutional Neural Network (Mask R-CNN) to detect macrophages in quantitative phase images, which were acquired by digital holographic microscopy (DHM), an interferometry-based variant of quantitative phase imaging (QPI). The Mask R-CNN deep learning architecture is capable to detect and segment single macrophage cells in quantitative phase images and allows to perform both tasks in a multi-stage process. Our results show that the combined detection and segmentation of cells through Mask R-CNN-based automated evaluation prospects a fast and robust screening in label-free high throughput microscopy.

Original languageEnglish
Title of host publicationLabel-free Biomedical Imaging and Sensing (LBIS) 2021
EditorsNatan T. Shaked, Oliver Hayden
PublisherSpie -- the Int Soc for Optical Engineering
ISBN (Electronic)9781510641457
DOIs
Publication statusPublished - 5 Mar 2021
EventLabel-free Biomedical Imaging and Sensing, LBIS 2021 - Virtual, Online, United States
Duration: 6 Mar 202111 Mar 2021

Publication series

NameProgress in Biomedical Optics and Imaging - Proceedings of SPIE
Volume11655
ISSN (Print)1605-7422

Conference

ConferenceLabel-free Biomedical Imaging and Sensing, LBIS 2021
CountryUnited States
CityVirtual, Online
Period6/03/2111/03/21

Keywords

  • Artificial intelligence
  • Deep learning
  • Digital holographic microscopy
  • Macrophages
  • Object detection
  • Quantitative phase imaging

Fingerprint

Dive into the research topics of 'Automated detection of macrophages in quantitative phase images by deep learning using a Mask Region-based Convolutional Neural Network'. Together they form a unique fingerprint.

Cite this