Unsupervised pathology detection in medical images using learning-based methods

Hristina Uzunova; Heinz Handels; Jan Ehrhardt

doi:10.1007/978-3-662-56537-7_30

Unsupervised pathology detection in medical images using learning-based methods

Hristina Uzunova^*, Heinz Handels, Jan Ehrhardt

^*Corresponding author for this work

Institute of Medical Informatics

4 Citations (Scopus)

Abstract

Detecting pathologies automatically is challenging because of their big variability. As the usual supervised machine learning approaches would only be able to detect one type of pathologies, in this work we pursue an unsupervised approach: learn the entire variability of healthy data and detect pathologies by their differences to the learned norm. Two methods have been developed based on this principle: A modified PatchMatch algorithm shows plausible results on contrasting brain tumors, but bad generalization ability for other types of data. A CVAE-based method on the other hand performs significantly better and ca. 17 times faster on the brain data and can be generalized to other pathologies, e.g. lung tumors. Not only is the achieved Dice coefficient of 0.55 comparable to other supervised methods on this data, moreover this method reliably detects different pathology types and needs no groundtruth.

Original language	English
Title of host publication	Bildverarbeitung für die Medizin 2018
Editors	A. Maier, T.M. Deserno, H. Handels, K.H. Maier-Hein, C. Palm, T. Tolxdorff
Number of pages	6
Publisher	Springer Vieweg, Berlin Heidelberg
Publication date	01.01.2018
Edition	211279
Pages	61-66
ISBN (Print)	978-3-662-56537-7
ISBN (Electronic)	978-3-662-56536-0
DOIs	https://doi.org/10.1007/978-3-662-56537-7_30
Publication status	Published - 01.01.2018
Event	Bildverarbeitung für die Medizin 2018 - Lehrstuhl für Mustererkennung, Erlangen, Germany Duration: 11.03.2018 → 13.03.2018 https://www.springer.com/us/book/9783662565360 http://www.bvm-workshop.org

UN SDGs

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

SDG 3 Good Health and Well-being
SDG 9 Industry, Innovation, and Infrastructure

Access to Document

10.1007/978-3-662-56537-7_30

Cite this

Uzunova, H., Handels, H., & Ehrhardt, J. (2018). Unsupervised pathology detection in medical images using learning-based methods. In A. Maier, T. M. Deserno, H. Handels, K. H. Maier-Hein, C. Palm, & T. Tolxdorff (Eds.), Bildverarbeitung für die Medizin 2018 (211279 ed., pp. 61-66). (Informatik aktuell). Springer Vieweg, Berlin Heidelberg. https://doi.org/10.1007/978-3-662-56537-7_30

@inproceedings{7626d378b91d44f197d338c5c610dbbc,

title = "Unsupervised pathology detection in medical images using learning-based methods",

abstract = "Detecting pathologies automatically is challenging because of their big variability. As the usual supervised machine learning approaches would only be able to detect one type of pathologies, in this work we pursue an unsupervised approach: learn the entire variability of healthy data and detect pathologies by their differences to the learned norm. Two methods have been developed based on this principle: A modified PatchMatch algorithm shows plausible results on contrasting brain tumors, but bad generalization ability for other types of data. A CVAE-based method on the other hand performs significantly better and ca. 17 times faster on the brain data and can be generalized to other pathologies, e.g. lung tumors. Not only is the achieved Dice coefficient of 0.55 comparable to other supervised methods on this data, moreover this method reliably detects different pathology types and needs no groundtruth.",

author = "Hristina Uzunova and Heinz Handels and Jan Ehrhardt",

year = "2018",

month = jan,

day = "1",

doi = "10.1007/978-3-662-56537-7\_30",

language = "English",

isbn = "978-3-662-56537-7",

series = "Informatik aktuell",

publisher = "Springer Vieweg, Berlin Heidelberg",

pages = "61--66",

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

booktitle = "Bildverarbeitung f{\"u}r die Medizin 2018",

edition = "211279",

note = "Bildverarbeitung f{\"u}r die Medizin 2018 ; Conference date: 11-03-2018 Through 13-03-2018",

url = "https://www.springer.com/us/book/9783662565360, http://www.bvm-workshop.org",

}

Uzunova, H , Handels, H & Ehrhardt, J 2018, Unsupervised pathology detection in medical images using learning-based methods. in A Maier, TM Deserno, H Handels, KH Maier-Hein, C Palm & T Tolxdorff (eds), Bildverarbeitung für die Medizin 2018. 211279 edn, Informatik aktuell, Springer Vieweg, Berlin Heidelberg, pp. 61-66, Bildverarbeitung für die Medizin 2018, Erlangen, Germany, 11.03.18. https://doi.org/10.1007/978-3-662-56537-7_30

Unsupervised pathology detection in medical images using learning-based methods. / Uzunova, Hristina ; Handels, Heinz ; Ehrhardt, Jan.
Bildverarbeitung für die Medizin 2018. ed. / A. Maier; T.M. Deserno; H. Handels; K.H. Maier-Hein; C. Palm; T. Tolxdorff. 211279. ed. Springer Vieweg, Berlin Heidelberg, 2018. p. 61-66 (Informatik aktuell).

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

TY - GEN

T1 - Unsupervised pathology detection in medical images using learning-based methods

AU - Uzunova, Hristina

AU - Handels, Heinz

AU - Ehrhardt, Jan

PY - 2018/1/1

Y1 - 2018/1/1

N2 - Detecting pathologies automatically is challenging because of their big variability. As the usual supervised machine learning approaches would only be able to detect one type of pathologies, in this work we pursue an unsupervised approach: learn the entire variability of healthy data and detect pathologies by their differences to the learned norm. Two methods have been developed based on this principle: A modified PatchMatch algorithm shows plausible results on contrasting brain tumors, but bad generalization ability for other types of data. A CVAE-based method on the other hand performs significantly better and ca. 17 times faster on the brain data and can be generalized to other pathologies, e.g. lung tumors. Not only is the achieved Dice coefficient of 0.55 comparable to other supervised methods on this data, moreover this method reliably detects different pathology types and needs no groundtruth.

AB - Detecting pathologies automatically is challenging because of their big variability. As the usual supervised machine learning approaches would only be able to detect one type of pathologies, in this work we pursue an unsupervised approach: learn the entire variability of healthy data and detect pathologies by their differences to the learned norm. Two methods have been developed based on this principle: A modified PatchMatch algorithm shows plausible results on contrasting brain tumors, but bad generalization ability for other types of data. A CVAE-based method on the other hand performs significantly better and ca. 17 times faster on the brain data and can be generalized to other pathologies, e.g. lung tumors. Not only is the achieved Dice coefficient of 0.55 comparable to other supervised methods on this data, moreover this method reliably detects different pathology types and needs no groundtruth.

UR - https://www.scopus.com/pages/publications/85043340170

U2 - 10.1007/978-3-662-56537-7_30

DO - 10.1007/978-3-662-56537-7_30

M3 - Conference contribution

AN - SCOPUS:85043340170

SN - 978-3-662-56537-7

T3 - Informatik aktuell

SP - 61

EP - 66

BT - Bildverarbeitung für die Medizin 2018

A2 - Maier, A.

A2 - Deserno, T.M.

A2 - Handels, H.

A2 - Maier-Hein, K.H.

A2 - Palm, C.

A2 - Tolxdorff, T.

PB - Springer Vieweg, Berlin Heidelberg

T2 - Bildverarbeitung für die Medizin 2018

Y2 - 11 March 2018 through 13 March 2018

ER -

Unsupervised pathology detection in medical images using learning-based methods

Abstract

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this