Table 3.

Precision, recall, and F1-scores for five different classifiers for FBA and fwcDTI metrics reported using five-fold cross-validation in HIV+ subjects only.

Classifiers	Metrics	Precision	Recall	F1-score
LDA	FBA	0.62 (0.067)	0.60 (0.054)	0.58 (0.054)
	fwcDTI	0.50 (0.045)	0.43 (0.049)	0.46 (0.049)
Random forest	FBA	0.62 (0.076)	0.77 (0.045)	0.76 (0.049)
	fwcDTI	0.53 (0.040)	0.53 (0.045)	0.53 (0.045)
naïve Bayes	FBA	0.62 (0.067)	0.58 (0.044)	0.55 (0.045)
	fwcDTI	0.38 (0.045)	0.38 (0.049)	0.38 (0.045)
AdaBoost	FBA	0.80 (0.049)	0.77 (0.045)	0.76 (0.049)
	fwcDTI	0.48 (0.058)	0.46 (0.049)	0.47 (0.049)

Scores reported as weighted average across three classes. FBA, fixel-based analysis; fwcDTI, free water corrected diffusion tensor imaging; LDA, linear discriminant analysis. Values are reported as mean (SE). Highest performance is shown as bold.