Deep embeddings for novelty detection in myopathy

Philippe Burlina, Neil Joshi, Seth Billings, I. Jeng Wang, Jemima Albayda

Research output: Contribution to journalArticle

Abstract

We address the challenge of finding anomalies in ultrasound images via deep learning, specifically applying this to screening for myopathies and finding rare presentations of myopathic disease. Among myopathic diseases, this study focuses on the use case of myositis given the spectrum of muscle involvement seen in these inflammatory muscle diseases, as well as the potential for treatment. For this study, we have developed a fully annotated dataset (called “Myositis3K”) which includes 3586 images of eighty-nine individuals (35 control and 54 with myositis) acquired with informed consent. We approach this challenge as one of performing unsupervised novelty detection (ND), and use tools leveraging deep embeddings combined with several novelty scoring methods. We evaluated these various ND algorithms and compared their performance against human clinician performance, against other methods including supervised binary classification approaches, and against unsupervised novelty detection approaches using generative methods. Our best performing approach resulted in a (ROC) AUC (and 95% CI error margin) of 0.7192 (0.0164), which is a promising baseline for developing future clinical tools for unsupervised prescreening of myopathies.

Original languageEnglish (US)
Pages (from-to)46-53
Number of pages8
JournalComputers in Biology and Medicine
Volume105
DOIs
StatePublished - Feb 1 2019

Fingerprint

Myositis
Muscular Diseases
Muscle
Informed Consent
Area Under Curve
Screening
Research Design
Ultrasonics
Learning
Muscles

Keywords

  • Deep embeddings
  • Deep learning
  • Muscular diseases
  • Novelty detection

ASJC Scopus subject areas

  • Computer Science Applications
  • Health Informatics

Cite this

Deep embeddings for novelty detection in myopathy. / Burlina, Philippe; Joshi, Neil; Billings, Seth; Wang, I. Jeng; Albayda, Jemima.

In: Computers in Biology and Medicine, Vol. 105, 01.02.2019, p. 46-53.

Research output: Contribution to journalArticle

Burlina, Philippe ; Joshi, Neil ; Billings, Seth ; Wang, I. Jeng ; Albayda, Jemima. / Deep embeddings for novelty detection in myopathy. In: Computers in Biology and Medicine. 2019 ; Vol. 105. pp. 46-53.
@article{62a06560a355418ebf14a6995891352c,
title = "Deep embeddings for novelty detection in myopathy",
abstract = "We address the challenge of finding anomalies in ultrasound images via deep learning, specifically applying this to screening for myopathies and finding rare presentations of myopathic disease. Among myopathic diseases, this study focuses on the use case of myositis given the spectrum of muscle involvement seen in these inflammatory muscle diseases, as well as the potential for treatment. For this study, we have developed a fully annotated dataset (called “Myositis3K”) which includes 3586 images of eighty-nine individuals (35 control and 54 with myositis) acquired with informed consent. We approach this challenge as one of performing unsupervised novelty detection (ND), and use tools leveraging deep embeddings combined with several novelty scoring methods. We evaluated these various ND algorithms and compared their performance against human clinician performance, against other methods including supervised binary classification approaches, and against unsupervised novelty detection approaches using generative methods. Our best performing approach resulted in a (ROC) AUC (and 95{\%} CI error margin) of 0.7192 (0.0164), which is a promising baseline for developing future clinical tools for unsupervised prescreening of myopathies.",
keywords = "Deep embeddings, Deep learning, Muscular diseases, Novelty detection",
author = "Philippe Burlina and Neil Joshi and Seth Billings and Wang, {I. Jeng} and Jemima Albayda",
year = "2019",
month = "2",
day = "1",
doi = "10.1016/j.compbiomed.2018.12.006",
language = "English (US)",
volume = "105",
pages = "46--53",
journal = "Computers in Biology and Medicine",
issn = "0010-4825",
publisher = "Elsevier Limited",

}

TY - JOUR

T1 - Deep embeddings for novelty detection in myopathy

AU - Burlina, Philippe

AU - Joshi, Neil

AU - Billings, Seth

AU - Wang, I. Jeng

AU - Albayda, Jemima

PY - 2019/2/1

Y1 - 2019/2/1

N2 - We address the challenge of finding anomalies in ultrasound images via deep learning, specifically applying this to screening for myopathies and finding rare presentations of myopathic disease. Among myopathic diseases, this study focuses on the use case of myositis given the spectrum of muscle involvement seen in these inflammatory muscle diseases, as well as the potential for treatment. For this study, we have developed a fully annotated dataset (called “Myositis3K”) which includes 3586 images of eighty-nine individuals (35 control and 54 with myositis) acquired with informed consent. We approach this challenge as one of performing unsupervised novelty detection (ND), and use tools leveraging deep embeddings combined with several novelty scoring methods. We evaluated these various ND algorithms and compared their performance against human clinician performance, against other methods including supervised binary classification approaches, and against unsupervised novelty detection approaches using generative methods. Our best performing approach resulted in a (ROC) AUC (and 95% CI error margin) of 0.7192 (0.0164), which is a promising baseline for developing future clinical tools for unsupervised prescreening of myopathies.

AB - We address the challenge of finding anomalies in ultrasound images via deep learning, specifically applying this to screening for myopathies and finding rare presentations of myopathic disease. Among myopathic diseases, this study focuses on the use case of myositis given the spectrum of muscle involvement seen in these inflammatory muscle diseases, as well as the potential for treatment. For this study, we have developed a fully annotated dataset (called “Myositis3K”) which includes 3586 images of eighty-nine individuals (35 control and 54 with myositis) acquired with informed consent. We approach this challenge as one of performing unsupervised novelty detection (ND), and use tools leveraging deep embeddings combined with several novelty scoring methods. We evaluated these various ND algorithms and compared their performance against human clinician performance, against other methods including supervised binary classification approaches, and against unsupervised novelty detection approaches using generative methods. Our best performing approach resulted in a (ROC) AUC (and 95% CI error margin) of 0.7192 (0.0164), which is a promising baseline for developing future clinical tools for unsupervised prescreening of myopathies.

KW - Deep embeddings

KW - Deep learning

KW - Muscular diseases

KW - Novelty detection

UR - http://www.scopus.com/inward/record.url?scp=85058692832&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85058692832&partnerID=8YFLogxK

U2 - 10.1016/j.compbiomed.2018.12.006

DO - 10.1016/j.compbiomed.2018.12.006

M3 - Article

VL - 105

SP - 46

EP - 53

JO - Computers in Biology and Medicine

T2 - Computers in Biology and Medicine

JF - Computers in Biology and Medicine

SN - 0010-4825

ER -