Unsupervised deep novelty detection: Application to muscle ultrasound and myositis screening

P. Burlina, N. Joshi, S. Billings, I. J. Wang, J. Albayda

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

This study investigates unsupervised novelty detection (ND) for screening of rare myopathies and specifically myositis. To support this study we developed from the ground up a novel and fully annotated dataset consisting of 3586 images taken of eighty nine individuals obtained under informed consent during 2016-2017. We developed and compared performance for several ND methods leveraging deep feature embeddings, utilizing generative as well as discriminative deep learning approaches for embeddings, and using various novelty scores. We carried out several performance comparisons including with a clinician, supervised binary classification approaches, and a generative method, demonstrating that our best performing approach is competitive with human performance and other best of breed algorithms.

Original languageEnglish (US)
Title of host publicationISBI 2019 - 2019 IEEE International Symposium on Biomedical Imaging
PublisherIEEE Computer Society
Pages1910-1914
Number of pages5
ISBN (Electronic)9781538636411
DOIs
StatePublished - Apr 2019
Event16th IEEE International Symposium on Biomedical Imaging, ISBI 2019 - Venice, Italy
Duration: Apr 8 2019Apr 11 2019

Publication series

NameProceedings - International Symposium on Biomedical Imaging
Volume2019-April
ISSN (Print)1945-7928
ISSN (Electronic)1945-8452

Conference

Conference16th IEEE International Symposium on Biomedical Imaging, ISBI 2019
CountryItaly
CityVenice
Period4/8/194/11/19

Keywords

  • Deep embeddings
  • Myopathy
  • Myositis
  • Novelty detection
  • Unsupervised learning

ASJC Scopus subject areas

  • Biomedical Engineering
  • Radiology Nuclear Medicine and imaging

Fingerprint Dive into the research topics of 'Unsupervised deep novelty detection: Application to muscle ultrasound and myositis screening'. Together they form a unique fingerprint.

Cite this