Detail publikace

Registration of medical image sequences using auto-differentiation

VIČAR, T. JAKUBÍČEK, R. CHMELÍK, J. KOLÁŘ, R.

Originální název

Registration of medical image sequences using auto-differentiation

Typ

článek ve sborníku ve WoS nebo Scopus

Jazyk

angličtina

Originální abstrakt

This paper focuses on image registration using the automatic differentiation of deep learning frameworks. Specifically, a method for the registration of image sequences is proposed and tested on retinal video ophthalmoscopic data and brain DCE MR images. PyTorch auto-differentiation has been used as a core of an optimisation tool to find the optimal image transformation parameters. It allows us to easily design a loss function for our registration tasks. The image registration was achieved by simultaneous registration of all images using a global loss function without the need of the reference frame.

Klíčová slova

medical image registration; auto-differentiation; deep learning frameworks; gradient-based optimisation; video stabilisation

Autoři

VIČAR, T.; JAKUBÍČEK, R.; CHMELÍK, J.; KOLÁŘ, R.

Vydáno

21. 12. 2023

Nakladatel

Springer

ISBN

978-981-16-6774-9

Kniha

Medical Imaging and Computer-Aided Diagnosis: Proceedings of 2022 International Conference on Medical Imaging and Computer-Aided Diagnosis (MICAD 2022)

Edice

1

ISSN

1876-1100

Periodikum

Lecture notes in Electrical Engineering

Ročník

810

Stát

Spojené království Velké Británie a Severního Irska

Strany od

169

Strany do

178

Strany počet

10

URL

https://link.springer.com/chapter/10.1007/978-981-16-6775-6_15

BibTex

@inproceedings{BUT180041,
  author="Tomáš {Vičar} and Roman {Jakubíček} and Jiří {Chmelík} and Radim {Kolář}",
  title="Registration of medical image sequences using auto-differentiation",
  booktitle="Medical Imaging and Computer-Aided Diagnosis: Proceedings of 2022 International Conference on Medical Imaging and Computer-Aided Diagnosis (MICAD 2022)",
  year="2023",
  series="1",
  journal="Lecture notes in Electrical Engineering",
  volume="810",
  pages="169--178",
  publisher="Springer",
  doi="10.1007/978-981-16-6775-6\{_}15",
  isbn="978-981-16-6774-9",
  issn="1876-1100",
  url="https://link.springer.com/chapter/10.1007/978-981-16-6775-6_15"
}