Prediction of estimated risk for bipolar disorder using machine learning and structural MRI features

Research output: Contribution to journalResearch articleContributedpeer-review

Contributors

  • Pavol Mikolas - , Department of Psychiatry and Psychotherapy (Author)
  • Michael Marxen - , Department of Psychiatry and Psychotherapy (Author)
  • Philipp Riedel - , Department of Psychiatry and Psychotherapy (Author)
  • Kyra Broeckel - , TUD Dresden University of Technology (Author)
  • Julia Martini - , Department of Psychiatry and Psychotherapy (Author)
  • Fabian Huth - , Department of Psychiatry and Psychotherapy, TUD Dresden University of Technology (Author)
  • Christina Berndt - , Department of Psychiatry and Psychotherapy (Author)
  • Christoph Vogelbacher - , University of Marburg, Justus Liebig University Giessen (Author)
  • Andreas Jansen - , University of Marburg, Justus Liebig University Giessen (Author)
  • Tilo Kircher - , University of Marburg, Justus Liebig University Giessen (Author)
  • Irina Falkenberg - , University of Marburg, Justus Liebig University Giessen (Author)
  • Martin Lambert - , University Hospital Hamburg Eppendorf (Author)
  • Vivien Kraft - , University Hospital Hamburg Eppendorf (Author)
  • Gregor Leicht - , University Hospital Hamburg Eppendorf (Author)
  • Christoph Mulert - , University of Marburg, Justus Liebig University Giessen, University Hospital Hamburg Eppendorf (Author)
  • Andreas J. Fallgatter - , University of Tübingen (Author)
  • Thomas Ethofer - , University of Tübingen (Author)
  • Anne Rau - , University of Tübingen (Author)
  • Karolina Leopold - , Free University of Berlin (Author)
  • Andreas Bechdolf - , Free University of Berlin (Author)
  • Andreas Reif - , University Hospital Frankfurt (Author)
  • Silke Matura - , University Hospital Frankfurt (Author)
  • Felix Bermpohl - , Free University of Berlin (Author)
  • Jana Fiebig - , Free University of Berlin (Author)
  • Thomas Stamm - , Free University of Berlin (Author)
  • Christoph U. Correll - , Free University of Berlin, Northwell Health System (Author)
  • Georg Juckel - , Ruhr University Bochum (Author)
  • Vera Flasbeck - , Ruhr University Bochum (Author)
  • Philipp Ritter - , Department of Psychiatry and Psychotherapy (Author)
  • Michael Bauer - , Department of Psychiatry and Psychotherapy (Author)
  • Andrea Pfennig - , Department of Psychiatry and Psychotherapy (Author)

Abstract

BACKGROUND: Individuals with bipolar disorder are commonly correctly diagnosed a decade after symptom onset. Machine learning techniques may aid in early recognition and reduce the disease burden. As both individuals at risk and those with a manifest disease display structural brain markers, structural magnetic resonance imaging may provide relevant classification features.

METHODS: Following a pre-registered protocol, we trained linear support vector machine (SVM) to classify individuals according to their estimated risk for bipolar disorder using regional cortical thickness of help-seeking individuals from seven study sites ( N = 276). We estimated the risk using three state-of-the-art assessment instruments (BPSS-P, BARS, EPI bipolar).

RESULTS: For BPSS-P, SVM achieved a fair performance of Cohen's κ of 0.235 (95% CI 0.11-0.361) and a balanced accuracy of 63.1% (95% CI 55.9-70.3) in the 10-fold cross-validation. In the leave-one-site-out cross-validation, the model performed with a Cohen's κ of 0.128 (95% CI -0.069 to 0.325) and a balanced accuracy of 56.2% (95% CI 44.6-67.8). BARS and EPI bipolar could not be predicted. In post hoc analyses, regional surface area, subcortical volumes as well as hyperparameter optimization did not improve the performance.

CONCLUSIONS: Individuals at risk for bipolar disorder, as assessed by BPSS-P, display brain structural alterations that can be detected using machine learning. The achieved performance is comparable to previous studies which attempted to classify patients with manifest disease and healthy controls. Unlike previous studies of bipolar risk, our multicenter design permitted a leave-one-site-out cross-validation. Whole-brain cortical thickness seems to be superior to other structural brain features.

Details

Original languageEnglish
Pages (from-to)1-11
Number of pages11
JournalPsychological medicine
Volume54
Issue number2
Publication statusE-pub ahead of print - May 2023
Peer-reviewedYes

External IDs

PubMed 37212052
ORCID /0000-0001-8870-0041/work/142251369
ORCID /0000-0001-9298-2125/work/143074533
ORCID /0000-0002-2666-859X/work/146643981
ORCID /0000-0003-4286-5830/work/148143972
ORCID /0000-0002-3974-7115/work/148144898
Mendeley 88541217-fe6f-390a-b262-debe67660ad2
unpaywall 10.1017/s0033291723001319
ORCID /0000-0002-3415-5583/work/150329695
Scopus 85184302170

Keywords

Keywords

  • Diagnostic classification, Machine learning, Risk of bipolar disorder, structural MRI