Medizinische Universität Graz Austria/Österreich - Forschungsportal - Medical University of Graz
Gewählte Publikation:
Elst, C.
Machine Based Learning of Multidimensional Data in Bipolar Disorder - Choice of Methods and Pilot Results
Humanmedizin; [ Diplomarbeit ] Medizinische Universität Graz; 2022. pp. 109
[OPEN ACCESS]
FullText
- Autor*innen der Med Uni Graz:
- Betreuer*innen:
- Birner Armin
- Kreuzthaler Markus Eduard
- Altmetrics:
- Abstract:
- Bipolar disorder is a psychiatric disorder with a high impact on a patient’s normal societal function and is associated with an increased risk of repeated hospitalizations and an elevated lifetime risk for suicide. Delays in the correct diagnosis and to the start of an effective treatment are associated with poorer outcomes, but are common due to a very heterogeneous progression of the disorder. Techniques of machine learning can be used to aid in the diagnosis of bipolar disorder and shorten the time from the onset of symptoms to the beginning of treatment. To test this hypothesis, a de-identified data set of demographic information and the results of cognitive tests of 196 patients with bipolar disorder and 145 healthy controls was used to train and compare five different machine learning algorithms. The best performing algorithm was Logistic Regression, with a macro-average F1 score of 0.69 [95% CI 0.66 - 0.73]. After further optimization, a model with an improved macro-average F1 score of 0.75, a micro-average F1 score of 0.77 and an AUROC of 0.84 could be built. Based on this model, it was analyzed how much each single variable contributes to the classification, which resulted in the finding that a patient’s BMI and results of the Stroop test and the d2/d2-R test alone allow for a classification with equal performance. Using this data for clinical application results in an acceptable performance, but has not yet reached a state where it can sufficiently augment a diagnosis made by an experienced clinician. The focus of further research should be to identify variables with a high contribution to classification to further improve the performance of machine learning models in this context and to subsequently simplify the diagnosis of bipolar disorder.