Wen-Bin Luo^{1}, Jing-Ying Huang^{2,3}, Yung-Chin Hsu^{2}, and Wen-Yih Isaac Tseng^{2,4,5,6}

Although white matter tract microstructure has been implicated in treatment outcome of schizophrenia, its predictive capability on first-episode patients remains unknown. In the study, diffusion spectrum imaging (DSI) data were acquired from both chronic and first-episode patients, reconstructed by mean apparent propagator (MAP) MRI and analyzed with tract-based automatic analysis (TBAA). Stepwise statistical analysis was then performed to identify specific segments of white matter tracts that were significantly different between remitted and non-remitted chronic patients. We built a support vector classifier on the preprocessed data matrix. The resulting model yielded fair validation and test accuracy on chronic and first-episode patients, respectively.

- Chong HY, Teoh SL, Wu DB, Kotirum S, Chiou CF, Chaiyakunapruk N. Global economic burden of schizophrenia: a systematic review. Neuropsychiatr Dis Treat. 2016;12:357-73.
- Özarslan E, Koay CG, Shepherd TM, et al. Mean apparent propagator (MAP) MRI: a novel diffusion imaging method for mapping tissue microstructure. Neuroimage. 2013;78:16-32.
- Andreasen NC, Carpenter WT, Kane JM, Lasser RA, Marder SR, Weinberger DR. Remission in schizophrenia: proposed criteria and rationale for consensus. Am J Psychiatry. 2005;162(3):441-9.
- Chen YJ, Lo YC, Hsu YC, et al. Automatic whole brain tract-based analysis using predefined tracts in a diffusion spectrum imaging template and an accurate registration strategy. Hum Brain Mapp. 2015;36(9):3441-58.
- Pedregosa et al. (2011). Scikit-learn: Machine Learning in Python. JMLR 12, pp. 2825-2830.
- Mitelman, S. A., Newmark, R. E., Torosjan, Y., Chu, K. W., Brickman, A. M., Haznedar, M. M., ... & Buchsbaum, M. S. (2006). White matter fractional anisotropy and outcome in schizophrenia. Schizophrenia research, 87(1), 138-159.
- Mitelman, S. A., Torosjan, Y., Newmark, R. E., Schneiderman, J. S., Chu, K. W., Brickman, A. M., ... & Buchsbaum, M. S. (2007). Internal capsule, corpus callosum and long associative fibers in good and poor outcome schizophrenia: a diffusion tensor imaging survey. Schizophrenia research, 92(1), 211-224.

The cost function for
support vector classifier, where C is the penalty, w is the decision variable
(the normal vector of the separating hyperplane), y_{i} is the class to
which the i-th sample belongs, x_{i} is the observed values of the i-th
sample.

The p value maps of
stepwise ANCOVA. -log(p) is mapped from dark red to light yellow. All steps
with p value ≤ 0.01 are colored white. In each plot, the x-axis represents 100 steps
obtained from TBAA, and the y-axis represents 76 white matter tracts.
Contiguous segments are contiguous significant steps of length above the 95th
percentile.

The validation curves
on the training set (chronic patients). The x-axis is the range from 2^{-10}
to 2^{10} that the penalty parameter C can possible take on. The y-axis
denotes accuracy. The solid line indicates the mean of four accuracies obtained
from four-fold cross validation, while the shaded area denotes ±1 standard
deviation. The vertical red bar denotes the value of C at which the validation
curve peaks.

ROC and AUC. ROC
plots sensitivity against 1-specificity. The best cutoff value is proposed to
be the point (marked red) that yields the maximum sum of sensitivity and
specificity. An AUC of nearly 0.9 indicates fair performance of a model on
learning the training data and that the model is complex enough to achieve good
fit.

The confusion matrix
for evaluating the performance of the support vector classifier. Abbreviations:
R (remitted), NR (non-remitted), PPV (positive predictive value), NPV (negative
predictive value).