Abstract
Objective: Unipolar depression and depressive episodes of bipolar disorder have similar symptoms and their differential diagnosis is crucial because each disorder has different prognostic and therapeutic characteristics. The main aim of the current study was to investigate white matter alterations as measured by fractional anisotropy in individuals with bipolar disorder (BD) and unipolar depression (UD) using tract-based spatial statistics and to find out if these alterations can help to make a differential diagnosis between these two disorders.
Methods: Tract-based spatial statistics is a sensitive method of whole-brain analysis that relies on the voxel-based comparison. It uses nonlinear image transformation and permutation tests with correction for multiple comparisons. The study consisted of total number of 107 subjects; whom were diagnosed clinically at least by two different psychiatrists and their data were reviewed by another psychiatrist retrospectively. Whole-brain diffusion tensor images of 41 patients with bipolar disorder type 1, 43 patients with unipolar depression and 23 healthy controls were acquired using a 1.5 Tesla magnetic resonance imaging scanner. The results were analyzed with 1. Whole brain analysis, 2. Region of Interests (ROI) analysis followed by machine learning methods: Genetic Algorithm and Kernel Logistic Regression.
Results: Compared to controls, UD and BD subjects showed reduced FA in several white matter tracts (p<0.05). However the age range of clinical groups was wider. To eliminate errors due to this difference in age ranges, we eliminated individuals from clinical groups and equalized the age ranges with that of the control group. However, even after restricting the age range of UD and BD subjects group, the results remained the same.
As compared to UD group, BD group showed significant FA reductions (p<0.001, uncorrected) in the following white matter tracts: corticospinal tract, anterior thalamic radiation in the right hemisphere, and inferior longitudinal fasciculus in the left hemisphere. There were not any significant reductions in the UD group as compared to the BD group. Whole-brain analysis did not show significant group difference between patients diagnosed bipolar disorder and unipolar depression after statistical corrections were applied.
However, ROI analysis followed by machine learning showed that patients with unipolar depression and bipolar disorder could be discriminated with a classification accuracy of 85.83% using logistic regression method.
Conclusion: Due to the small number of study in the literature, which directly compared patients with unipolar depression and bipolar disorder, the current study aims to improve the understanding of the etiology and pathogenesis of bipolar disorder and unipolar depression. The results of the present study are consistent with the current understanding of bipolar disorder neurobiology. This may mean that fractional anisotropy values can be used as a biomarker to differentiate bipolar disorder from unipolar depression if further confirmed by larger studies.

Key words: bipolar disorder, unipolar depression, tract-based spatial statistics, white matter