Shallow fully-connected neural networks for ischemic stroke-lesion segmentation in MRI

Christian Lucas; Oskar Maier; Mattias P. Heinrich

doi:10.1007/978-3-662-54345-0_59

Shallow fully-connected neural networks for ischemic stroke-lesion segmentation in MRI

Christian Lucas, Oskar Maier, Mattias P. Heinrich

Institute of Medical Informatics

Graduate School for Computing in Medicine and Live Science

Abstract

Automatic image segmentation of stroke lesions could be of great importance for aiding the treatment decision. Convolutional neural networks obtain high accuracy for this task at the cost of prohibitive computational demand for time-sensitive clinical scenarios. In this work, we study the use of classical fully-connected neural networks (FC-NN) based on hand-crafted features, which achieve much shorter runtimes. We show that recent advances in optimization and regularization of deep learning can be successfully transferred to FC-NNs to improve the training process and achieve comparable accuracy to random decision forests.

Original language	English
Title of host publication	Bildverarbeitung für die Medizin 2017
Editors	K.H. Maier-Hein, T.M. Deserno, H. Handels, T. Tolxdorff
Number of pages	6
Publisher	Springer Vieweg, Berlin Heidelberg
Publication date	01.03.2017
Edition	1
Pages	261-266
ISBN (Print)	978-3-662-54344-3
ISBN (Electronic)	978-3-662-54345-0
DOIs	https://doi.org/10.1007/978-3-662-54345-0_59
Publication status	Published - 01.03.2017
Event	Bildverarbeitung für die Medizin 2017 - Heidelberg, Germany Duration: 12.03.2017 → 14.03.2017

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1007/978-3-662-54345-0_59

https://www.springer.com/de/book/9783662543443

Cite this

Lucas, C., Maier, O., & Heinrich, M. P. (2017). Shallow fully-connected neural networks for ischemic stroke-lesion segmentation in MRI. In K. H. Maier-Hein, T. M. Deserno, H. Handels, & T. Tolxdorff (Eds.), Bildverarbeitung für die Medizin 2017 (1 ed., pp. 261-266). (Inforamtik aktuell). Springer Vieweg, Berlin Heidelberg. https://doi.org/10.1007/978-3-662-54345-0_59

@inproceedings{7bfccb10507d4d43be978f38b3b664a1,

title = "Shallow fully-connected neural networks for ischemic stroke-lesion segmentation in MRI",

abstract = "Automatic image segmentation of stroke lesions could be of great importance for aiding the treatment decision. Convolutional neural networks obtain high accuracy for this task at the cost of prohibitive computational demand for time-sensitive clinical scenarios. In this work, we study the use of classical fully-connected neural networks (FC-NN) based on hand-crafted features, which achieve much shorter runtimes. We show that recent advances in optimization and regularization of deep learning can be successfully transferred to FC-NNs to improve the training process and achieve comparable accuracy to random decision forests.",

author = "Christian Lucas and Oskar Maier and Heinrich, {Mattias P.}",

year = "2017",

month = mar,

day = "1",

doi = "10.1007/978-3-662-54345-0_59",

language = "English",

isbn = "978-3-662-54344-3",

series = "Inforamtik aktuell",

publisher = "Springer Vieweg, Berlin Heidelberg",

pages = "261--266",

editor = "K.H. Maier-Hein and T.M. Deserno and H. Handels and T. Tolxdorff",

booktitle = "Bildverarbeitung f{\"u}r die Medizin 2017",

edition = "1",

note = "Bildverarbeitung f{\"u}r die Medizin 2017<br/> ; Conference date: 12-03-2017 Through 14-03-2017",

}

Lucas, C, Maier, O & Heinrich, MP 2017, Shallow fully-connected neural networks for ischemic stroke-lesion segmentation in MRI. in KH Maier-Hein, TM Deserno, H Handels & T Tolxdorff (eds), Bildverarbeitung für die Medizin 2017 . 1 edn, Inforamtik aktuell, Springer Vieweg, Berlin Heidelberg, pp. 261-266, Bildverarbeitung für die Medizin 2017
, Heidelberg, Germany, 12.03.17. https://doi.org/10.1007/978-3-662-54345-0_59

Shallow fully-connected neural networks for ischemic stroke-lesion segmentation in MRI. / Lucas, Christian; Maier, Oskar; Heinrich, Mattias P.
Bildverarbeitung für die Medizin 2017 . ed. / K.H. Maier-Hein; T.M. Deserno; H. Handels; T. Tolxdorff. 1. ed. Springer Vieweg, Berlin Heidelberg, 2017. p. 261-266 (Inforamtik aktuell).

Research output: Chapters in Books/Reports/Conference Proceedings › Conference contribution › peer-review

TY - GEN

T1 - Shallow fully-connected neural networks for ischemic stroke-lesion segmentation in MRI

AU - Lucas, Christian

AU - Maier, Oskar

AU - Heinrich, Mattias P.

PY - 2017/3/1

Y1 - 2017/3/1

N2 - Automatic image segmentation of stroke lesions could be of great importance for aiding the treatment decision. Convolutional neural networks obtain high accuracy for this task at the cost of prohibitive computational demand for time-sensitive clinical scenarios. In this work, we study the use of classical fully-connected neural networks (FC-NN) based on hand-crafted features, which achieve much shorter runtimes. We show that recent advances in optimization and regularization of deep learning can be successfully transferred to FC-NNs to improve the training process and achieve comparable accuracy to random decision forests.

AB - Automatic image segmentation of stroke lesions could be of great importance for aiding the treatment decision. Convolutional neural networks obtain high accuracy for this task at the cost of prohibitive computational demand for time-sensitive clinical scenarios. In this work, we study the use of classical fully-connected neural networks (FC-NN) based on hand-crafted features, which achieve much shorter runtimes. We show that recent advances in optimization and regularization of deep learning can be successfully transferred to FC-NNs to improve the training process and achieve comparable accuracy to random decision forests.

U2 - 10.1007/978-3-662-54345-0_59

DO - 10.1007/978-3-662-54345-0_59

M3 - Conference contribution

SN - 978-3-662-54344-3

T3 - Inforamtik aktuell

SP - 261

EP - 266

BT - Bildverarbeitung für die Medizin 2017

A2 - Maier-Hein, K.H.

A2 - Deserno, T.M.

A2 - Handels, H.

A2 - Tolxdorff, T.

PB - Springer Vieweg, Berlin Heidelberg

T2 - Bildverarbeitung für die Medizin 2017<br/>

Y2 - 12 March 2017 through 14 March 2017

ER -

Shallow fully-connected neural networks for ischemic stroke-lesion segmentation in MRI

Abstract

UN SDGs

Access to Document

Fingerprint

Cite this