Visual Analytics for Concept Exploration in Subspaces of Patient Groups: Making Sense of Complex Datasets with the Doctor-in-the-Loop

Michael Hund, Dominic Boehm, Werner Josef Sturm, Michael Sedlmair, Tobias Schreck, Torsten Ullrich, Daniel Keim, Ljiljana Majnaric, Andreas Holzinger

Publikation: Beitrag in einer FachzeitschriftArtikelForschungBegutachtung

Abstract

Sowohl Klinikerinnen und Kliniker als auch Forscherinnen und Forscher sind im biomedizinischen Bereich vermehrt mit hochdimensionalen und komplexen Patientinnendaten und Patientendaten konfrontiert. Durch den im hochdimensionalen Raum auftretenden sogenannten „curse of dimensionality“ stellen sich extrem schwierige Herausforderungen im maschinellen Lernen um Wissen in solchen Daten zu entdecken. Dabei sind insbesondere unwichtige, korrelierende und sich widersprechende Dimensionen gerade jene Einflüsse die unter anderen Ähnlichkeitsdefinitionen zwischen Datenpunkten die größten Einflüsse bewirken und dadurch z.B. das Clusteringergebnis verfälschen. Ein weiteres Muster welches ebenfalls durch den „curse of dimensionality“ beeinflusst wird ist die Korrelation zwischen einem Patientinnenzustand bzw. Patientenzustand und einem Therapieergebnis in bestimmten Kombinationen von Dimensionen (=Subspace). Die hochdimensionalen Daten müssen in eine niedrigere Anzahl von relevanten Dimensionen projiziert werden um es einem Domainexperten bzw. einer Domainexpertin (expert-in-the-loop) ermöglichen, erfolgreiche Korrelationen zu erkennen. Zusammen mit der Keim-Gruppe der Universität Konstanz, konnten wir einige Experimente durchführen, um die Nützlichkeit und das zukünftige Potential interaktiver Subspaceanalysetechniken (SubVis) mit einem human-in-the-loop zu zeigen. Die Experimente wurden mit real-world Datensätzen aus dem Universitätsklinikum Osijek (Kroatien) durchgeführt.
Spracheenglisch
Seiten233-247
FachzeitschriftBrain Informatics
Jahrgang3
Ausgabennummer4
DOIs
StatusVeröffentlicht - 17 Nov 2016

Fingerprint

Medicine

Schlagwörter

  • Maschinelles Lernen
  • interaktives maschinelles Lernen
  • interactive visuelle Analyse
  • Medizinische Informatik

ASJC Scopus subject areas

  • Artificial intelligence

Fields of Expertise

  • Information, Communication & Computing

Treatment code (Nähere Zuordnung)

  • Basic - Fundamental (Grundlagenforschung)

Dies zitieren

Visual Analytics for Concept Exploration in Subspaces of Patient Groups: Making Sense of Complex Datasets with the Doctor-in-the-Loop. / Hund, Michael; Boehm, Dominic; Sturm, Werner Josef; Sedlmair, Michael; Schreck, Tobias; Ullrich, Torsten; Keim, Daniel; Majnaric, Ljiljana; Holzinger, Andreas.

in: Brain Informatics, Jahrgang 3, Nr. 4, 17.11.2016, S. 233-247.

Publikation: Beitrag in einer FachzeitschriftArtikelForschungBegutachtung

Hund, Michael ; Boehm, Dominic ; Sturm, Werner Josef ; Sedlmair, Michael ; Schreck, Tobias ; Ullrich, Torsten ; Keim, Daniel ; Majnaric, Ljiljana ; Holzinger, Andreas. / Visual Analytics for Concept Exploration in Subspaces of Patient Groups: Making Sense of Complex Datasets with the Doctor-in-the-Loop. in: Brain Informatics. 2016 ; Jahrgang 3, Nr. 4. S. 233-247.
@article{ba2f8ca823b044d788f38e6c23416749,
title = "Visual Analytics for Concept Exploration in Subspaces of Patient Groups: Making Sense of Complex Datasets with the Doctor-in-the-Loop",
abstract = "Medical doctors and researchers in bio-medicine are increasingly confronted with complex patient data, posing new and difficult analysis challenges. These data are often comprising high-dimensional descriptions of patient conditions and measurements on the success of certain therapies. An important analysis question in such data is to compare and correlate patient conditions and therapy results along with combinations of dimensions. As the number of dimensions is often very large, one needs to map them to a smaller number of relevant dimensions to be more amenable for expert analysis. This is because irrelevant, redundant, and conflicting dimensions can negatively affect effectiveness and efficiency of the analytic process (the so-called curse of dimensionality). However, the possible mappings from high- to low-dimensional spaces are ambiguous. For example, the similarity between patients may change by considering different combinations of relevant dimensions (subspaces). We demonstrate the potential of subspace analysis for the interpretation of high-dimensional medical data. Specifically, we present SubVIS, an interactive tool to visually explore subspace clusters from different perspectives, introduce a novel analysis workflow, and discuss future directions for high-dimensional (medical) data analysis and its visual exploration. We apply the presented workflow to a real-world dataset from the medical domain and show its usefulness with a domain expert evaluation.",
keywords = "Knowledge Discovery, Machine Learning, Health Informatics, Subspace Clustering, Visual Analytics, Maschinelles Lernen, interaktives maschinelles Lernen, interactive visuelle Analyse, Medizinische Informatik",
author = "Michael Hund and Dominic Boehm and Sturm, {Werner Josef} and Michael Sedlmair and Tobias Schreck and Torsten Ullrich and Daniel Keim and Ljiljana Majnaric and Andreas Holzinger",
year = "2016",
month = "11",
day = "17",
doi = "10.1007/s40708-016-0043-5",
language = "English",
volume = "3",
pages = "233--247",
journal = "Brain Informatics",
issn = "2198-4018",
publisher = "SpringerOpen",
number = "4",

}

TY - JOUR

T1 - Visual Analytics for Concept Exploration in Subspaces of Patient Groups: Making Sense of Complex Datasets with the Doctor-in-the-Loop

AU - Hund, Michael

AU - Boehm, Dominic

AU - Sturm, Werner Josef

AU - Sedlmair, Michael

AU - Schreck, Tobias

AU - Ullrich, Torsten

AU - Keim, Daniel

AU - Majnaric, Ljiljana

AU - Holzinger, Andreas

PY - 2016/11/17

Y1 - 2016/11/17

N2 - Medical doctors and researchers in bio-medicine are increasingly confronted with complex patient data, posing new and difficult analysis challenges. These data are often comprising high-dimensional descriptions of patient conditions and measurements on the success of certain therapies. An important analysis question in such data is to compare and correlate patient conditions and therapy results along with combinations of dimensions. As the number of dimensions is often very large, one needs to map them to a smaller number of relevant dimensions to be more amenable for expert analysis. This is because irrelevant, redundant, and conflicting dimensions can negatively affect effectiveness and efficiency of the analytic process (the so-called curse of dimensionality). However, the possible mappings from high- to low-dimensional spaces are ambiguous. For example, the similarity between patients may change by considering different combinations of relevant dimensions (subspaces). We demonstrate the potential of subspace analysis for the interpretation of high-dimensional medical data. Specifically, we present SubVIS, an interactive tool to visually explore subspace clusters from different perspectives, introduce a novel analysis workflow, and discuss future directions for high-dimensional (medical) data analysis and its visual exploration. We apply the presented workflow to a real-world dataset from the medical domain and show its usefulness with a domain expert evaluation.

AB - Medical doctors and researchers in bio-medicine are increasingly confronted with complex patient data, posing new and difficult analysis challenges. These data are often comprising high-dimensional descriptions of patient conditions and measurements on the success of certain therapies. An important analysis question in such data is to compare and correlate patient conditions and therapy results along with combinations of dimensions. As the number of dimensions is often very large, one needs to map them to a smaller number of relevant dimensions to be more amenable for expert analysis. This is because irrelevant, redundant, and conflicting dimensions can negatively affect effectiveness and efficiency of the analytic process (the so-called curse of dimensionality). However, the possible mappings from high- to low-dimensional spaces are ambiguous. For example, the similarity between patients may change by considering different combinations of relevant dimensions (subspaces). We demonstrate the potential of subspace analysis for the interpretation of high-dimensional medical data. Specifically, we present SubVIS, an interactive tool to visually explore subspace clusters from different perspectives, introduce a novel analysis workflow, and discuss future directions for high-dimensional (medical) data analysis and its visual exploration. We apply the presented workflow to a real-world dataset from the medical domain and show its usefulness with a domain expert evaluation.

KW - Knowledge Discovery

KW - Machine Learning

KW - Health Informatics

KW - Subspace Clustering

KW - Visual Analytics

KW - Maschinelles Lernen

KW - interaktives maschinelles Lernen

KW - interactive visuelle Analyse

KW - Medizinische Informatik

U2 - 10.1007/s40708-016-0043-5

DO - 10.1007/s40708-016-0043-5

M3 - Article

VL - 3

SP - 233

EP - 247

JO - Brain Informatics

T2 - Brain Informatics

JF - Brain Informatics

SN - 2198-4018

IS - 4

ER -